Deep Networks: Overview
From Ufldl
(→Availability of data) |
|||
Line 52: | Line 52: | ||
models. The fine-tuning step can be done only using labeled data. In | models. The fine-tuning step can be done only using labeled data. In | ||
practice, by using unlabeled data to learn a good initial value for the weights in all layers | practice, by using unlabeled data to learn a good initial value for the weights in all layers | ||
- | + | <math>\textstyle W^{(l)}</math>, we usually get much better classifiers | |
after fine-tuning. | after fine-tuning. | ||